Tesis doctoral de Fernando García Vílchez
This work is devoted to present a new coding system based on the ccsds recommendation for image data compression (idc). The new coding system adds to the idc recommendation several important featured advantages: it allows any number of spatial wavelet decomposition levels; it provides scalability by quality, position, resolution, and component; and it supports multi/hyper/ultraspectral data coding, allowing to perform also an spectral decorrelation if requested. thanks to the generalization of the number of spatial wavelet decomposition levels, the compression performance is improved for all the target bit rates, specially for medium and high compression ratios. scalability is added by embedding headers. Thanks to the new file syntax, an encoded stream is scalable in terms of: quality, position, resolution, and component. Thus, a region of the image can be recovered without needing to decode the whole encoded stream. Thus, both the computational cost and the memory load of the end-user application are drastically reduced. In addition, it is possible to interactive transmit an encoded stream using a communication protocol such as jpip. results illustrate that the embedded headers do not produce header overhead. We also note that, with our approach, it is feasible to provide some of the proposed extensions to an already encoded compliant code-stream: the bit plane encoder (bpe) needs to be inverted, then the encoded streams may be reordered, and, at the same time, the embedded headers are inserted; but notice that the bpe is not performed again, and that the transform stage is left untouched. The resulting encoded stream has now all four different types of scalability. remote sensing applications have to deal with a huge volume ofmulti/hyper/ultraspectral data captured by sensors. The support to this kind of data is, therefore, essential. On the one hand, the idc recommendation is oriented to monoband compression: each spectral band is encoded separately, thus, if multiple bands are to be encoded, the target rate must be fixed for each band before the encoding process begins. On the other hand, our proposal is able to encode multiple bands together, possibly after a decorrelation in the spectral domain, thus, the bit allocation for each band can be properly fit and the compression performance is improved with respect to the idc recommendation. If an spectral decorrelator is requested, the significance of the spectral bands is not, in general, equally distributed. consequently, being able to set properly the target rate for each band becomes specifically important when an spectral decorrelator is employed. as a consequence of the proposed extensions, the compression performance is notably improved with respect to the idc recommendation for a large variety of remote sensing images, both monoband and multi/hyper/ultraspectral images. In addition, the new file syntax provides scalability. Thus, an encoded image can be interactive decoded and transmitted. when compared to the jpeg2000 standard, the reference technique, results suggest that our proposal, being less complex than jpeg2000, is competitive offering both similar compression results, and alike interactive decoding and transmission features.
Datos académicos de la tesis doctoral «Extending the ccsds recommendation for image data compression for remote sensing scenarios«
- Título de la tesis: Extending the ccsds recommendation for image data compression for remote sensing scenarios
- Autor: Fernando García Vílchez
- Universidad: Autónoma de barcelona
- Fecha de lectura de la tesis: 10/12/2009
Dirección y tribunal
- Director de la tesis
- Joan Serra Sagrista
- Tribunal
- Presidente del tribunal: josep Rifa coma
- david Megías jiménez (vocal)
- (vocal)
- (vocal)